Suggestions
Greg Linden
mostly retired internet relic
Greg Linden is a seasoned professional with nearly 30 years of experience in artificial intelligence, personalization, and recommendations, bridging the gap between research and product development.
Throughout his career, Greg has held prominent positions at tech giants like Amazon, Google, and Microsoft, steering large teams and projects to massive scales.
He is renowned for inventing Amazon's highly successful item-to-item collaborative filtering algorithm, which has become widely adopted in the industry.
Greg Linden's expertise spans various domains, including data science, AI experimentation, research, and software engineering, with significant contributions to patents and academic publications.
He holds a Master's degree in Computer Science from the University of Washington and a Master's in Management from Stanford University Graduate School of Business, complemented by a Bachelor's in Computer Science & Political Science from UC San Diego.
In addition to his corporate roles, Greg has also founded his own startups like Geeky Ventures and Findory, while providing advisory services to startups that have seen successful acquisition.
His diverse background and experience make him a valuable asset with an in-depth understanding of the tech industry's evolution and future trends.
Highlights
I posted a long excerpt from my book, "Algorithms and Misinformation", about computational propaganda, the manipulation of social media using a bunch of accounts. I don't think most people are aware of how successful disinformation has been in recent years. For example, at a point a few years ago, all of the top groups on Facebook used by Christians in the US were controlled by foreign adversaries. Through that and other efforts, about half of US voters saw and may have been influenced by at least some foreign propaganda. Propaganda is easier than you might think. Because of how trending and other algorithms work, just controlling a few hundred accounts, and often less, is enough to make social media sites feature some of your content. A few bots, sockpuppets, and shills go a long way. This post on my blog quotes the hard work of many researchers and journalists. It summarizes their work around the theme of fixing the algorithms that control what we all see on the internet. I hope it is useful to you.
The changes at Twitter to promote paid accounts blur the distinction between a recommender engine and an ad engine. Short-term optimizations like this are tempting to do, but have a long-term cost, distrust of the recommendations and reduced retention.
The internet is full of examples of optimizations that are short-term revenue wins but hurt long-term revenue and profitability. Double the number of ads you show. Recommend profitable items instead of the most relevant and useful items. Allowing scams or disinformation because they boost engagement. Pop up or pop over ads. What these have in common is that they bump metrics this quarter but hurt the long-term success of the company.
Recommender engines are not ad engines. Recommendations should be about helping people discover something new and useful, keeping customers coming back again and again.
